基于数据挖掘的区域暂态电压稳定评估

Data Mining Based Regional Transient Voltage Stability Assessment

针对区域暂态电压稳定评估相关理论还不完善、工程判据可靠性不足等问题,提出了基于数据挖掘的区域暂态电压稳定评估方法,构建了综合考虑单点负荷稳定与多点电压相互影响的2层评价框架。利用节点稳定度量指标及基于辨识的电压无功灵敏度矩阵提取网络原始特征。面对区域暂态电压失稳尚无可靠界定标准的难题,采用基于约束的半监督学习方式对数据集进行可靠分类。基于决策树算法建立逐步更新的分类模型,生成区域暂态电压稳定判据,通过模型挖掘出有关电压分区、代表节点的内在规律。EPRI 36节点系统上的仿真结果证明了评估方案的有效性,以及分类评估模型的适应性和准确性。

In allusion to the imperfection of the theory related to regional transient voltage stability assessment and insufficient reliability of engineering criteria, a data mining based method to assess regional transient voltage stability is proposed and a two-layer assessment framework, in which the interaction between single bus load stability and multi bus voltages is synthetically considered, is constructed. Utilizing the measure index of nodal stability and the identification based voltage/reactive power sensitivity matrix the original features of the power network is extracted. To cope with the difficult problem that there is not yet reliable standard to delimitate the regional transient voltage instability, a constraint based semi-supervised learning method is used to reliably classify and label the data sets. Based on decision tree algorithm a step-by-step updated classification model is established to generate the criterion of regional transient voltage stability, and by use of this classification model the inherent law related to the voltage partition and representative buses can be mined. The validity of the proposed assessment scheme as well as the adaptability and the accuracy of the classification model are verified by simulation results of EPRI 36-bus system.